Project

TODO

passar tabelas pra csv(todas?), rename blank hashtag/media, juntar valores de reply pra cluster? par(mfrow=c(1,2))

For R beginners

New chunk Ctrl+Alt+I

Execute chunk Ctrl+Shift+Enter

Execute all chunks Ctrl+Alt+R

HTML preview Ctrl+Shift+K

Library preparations

library(readr)

Attaching package: ‘readr’

The following object is masked from ‘package:koRpus’:

    tokenize

library(dplyr)

Attaching package: ‘dplyr’

The following objects are masked from ‘package:stats’:

    filter, lag

The following objects are masked from ‘package:base’:

    intersect, setdiff, setequal, union

library(tidyverse)

── Attaching core tidyverse packages ───────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse 2.0.0 ──
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.1     ✔ tibble    3.2.1
✔ lubridate 1.9.3     ✔ tidyr     1.3.1
✔ purrr     1.0.2     ── Conflicts ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
✖ ggplot2::annotate() masks NLP::annotate()
✖ dplyr::filter()     masks stats::filter()
✖ dplyr::lag()        masks stats::lag()
✖ readr::tokenize()   masks koRpus::tokenize()
ℹ Use the ]8;;http://conflicted.r-lib.org/conflicted package]8;; to force all conflicts to become errors

library(ggplot2)
library(reshape2)

Attaching package: ‘reshape2’

The following object is masked from ‘package:tidyr’:

    smiths

library(stats)

Data Import

data <- read.csv("~/4year/2semester/dtII/CSVs/HEIs.csv",
                 colClasses = c(tweet_id = "character"))

# Modifying created_at type so that attribute can be used more easily 
data$created_at <- as.POSIXct(data$created_at,
                              format= "%Y-%m-%dT%H:%M:%S", tz="UTC")

#View(data)
summary(data)

      id              tweet_id             text               type           bookmark_count    favorite_count     retweet_count      reply_count      
 Length:11728       Length:11728       Length:11728       Length:11728       Min.   :  0.000   Min.   :    0.00   Min.   :   0.00   Min.   :   0.000  
 Class :character   Class :character   Class :character   Class :character   1st Qu.:  0.000   1st Qu.:    7.00   1st Qu.:   2.00   1st Qu.:   0.000  
 Mode  :character   Mode  :character   Mode  :character   Mode  :character   Median :  0.000   Median :   20.00   Median :   5.00   Median :   1.000  
                                                                             Mean   :  1.543   Mean   :   60.67   Mean   :  10.62   Mean   :   3.888  
                                                                             3rd Qu.:  1.000   3rd Qu.:   57.00   3rd Qu.:  11.00   3rd Qu.:   3.000  
                                                                             Max.   :418.000   Max.   :41655.00   Max.   :4214.00   Max.   :2317.000  
                                                                                                                                                      
   view_count        created_at                       hashtags             urls            media_type         media_urls       
 Min.   :      5   Min.   :2022-08-01 03:05:11.00   Length:11728       Length:11728       Length:11728       Length:11728      
 1st Qu.:   2643   1st Qu.:2022-10-19 12:56:27.00   Class :character   Class :character   Class :character   Class :character  
 Median :   6240   Median :2023-01-29 08:26:30.00   Mode  :character   Mode  :character   Mode  :character   Mode  :character  
 Mean   :  14182   Mean   :2023-01-30 07:39:34.96                                                                              
 3rd Qu.:  16058   3rd Qu.:2023-05-05 14:16:43.25                                                                              
 Max.   :7604544   Max.   :2023-08-31 20:50:01.00                                                                              
 NA's   :4840                                                                                                                  

Initial Data Preparation

# Count of how many entries each HEI has
number_interactions <- data %>%
              group_by(id) %>% summarise(count = n())

number_interactions

Since complutense only has 1 entry we can’t learn anything from it, so we removed it

data <- data[data$id != "complutense.csv", ]

Visualization of number all posts, just tweets and just replies

number_posts <- data %>%
              group_by(id) %>% summarise(count = n())

number_tweets <- data[data$type == "Tweet", ] %>%
              group_by(id) %>% summarise(count = n())

number_replies <- data[data$type == "Reply", ] %>%
              group_by(id) %>% summarise(count = n())

print(number_posts)

print(number_tweets)

print(number_replies)

Calculating the percentage of tweets and replies based on all posts

# Merging the counts of tweets (count.y) and replies (count) with the count of posts (count.x)
data_ratio <- merge(number_posts, number_tweets, by = "id", all = TRUE)
data_ratio <- merge(data_ratio, number_replies, by = "id", all = TRUE)


data_ratio$percentage_tweets <- (data_ratio$count.y / data_ratio$count.x) * 100
data_ratio$percentage_replies <- (data_ratio$count / data_ratio$count.x) * 100

data_ratio <- data_ratio[, c("id", "percentage_tweets", "percentage_replies")]

print(data_ratio)

NA removal

Function to visualize the number of NAs in all columns

na_count <- function(){
  # Counting the number of NA values for each column
  na_count <- colSums(is.na(data))
  
  # Creating a new data frame with the NA counts
  na_counts_table <- data.frame(Column = names(na_count), NA_Count = na_count)
  
  print(na_counts_table)
}

Calculations of view, favourite, retweet and reply percentiles and visualization of NAs in all columns

data <- data %>%
  group_by(id) %>%
  mutate(view_percentile = ntile(view_count, 100),
         favorite_percentile = ntile(favorite_count, 100),
         retweet_percentile = ntile(retweet_count, 100),
         reply_percentile = ntile(reply_count, 100)) %>%
  rowwise() %>%
  mutate(avg_percentile = mean(c(view_percentile, favorite_percentile, retweet_percentile, reply_percentile), na.rm = TRUE))

na_count()

data_percentile <- data[, c("id", "view_percentile", "favorite_percentile", "retweet_percentile", "reply_percentile", "avg_percentile")]

print(data_percentile)

Calculation of the maximum number of views for each HEI

max_view_counts <- tapply(data$view_count, data$id, max, na.rm = TRUE)

print(max_view_counts)

      duke.csv       epfl.csv        goe.csv    harvard.csv  leicester.csv manchester.csv        mit.csv         sb.csv   stanford.csv    trinity.csv 
        307969         105095          15455        2982704          47838         317086        7604544         607498         222593         205333 
        wv.csv       yale.csv 
        109265         143108 

Removal of NAs

# From view count
data$view_count <- ifelse(
  is.na(data$view_count),
  round(max_view_counts[data$id] * (data$avg_percentile / 100)),
  data$view_count)

# From view percentile
data$view_percentile <- ifelse(
  is.na(data$view_percentile),
  data$avg_percentile,
  data$view_percentile)

Visualization of NAs in all columns

na_count()

For now we’ll be only looking at tweets

data_tweets <- data[data$type == "Tweet", ]

data_tweets

Function to calculate average posts

average_tweets <- function(timeframe = "days"){
  # Calculation of the timeframe between earliest and latest post for each HEI
  date_range <- data_tweets %>%
    group_by(id) %>%
    summarise(min_date = min(created_at),
              max_date = max(created_at)) %>%
    mutate(num_days = as.numeric(difftime(max_date, min_date, units = timeframe)))
  
  # Naming the column respecting the timeframe
  column_name <- paste0("avg_tweets_per_", timeframe)
  
  # Calculation of the number of tweets per day for each HEI
  tweets_per_timeframe <- number_tweets %>%
    left_join(date_range, by = "id") %>%
    mutate(!!column_name := count / num_days)
  
  print(tweets_per_timeframe)
  return(tweets_per_timeframe)
}

tweets_per_day <- average_tweets()

tweets_per_week <- average_tweets(timeframe = "weeks")

Plot for the average number of tweets per day for each HEI

barplot(tweets_per_day$avg_tweets_per_days,
        names.arg = tweets_per_day$id,
        main = "Average Tweets per Day",
        xlab = "HEI",
        ylab = "Average Number of Tweets",
        ylim = c(0, max(tweets_per_day$avg_tweets_per_days) + 1),
        las = 2,
        col = "#3498DB")

# Adding text labels over each bar and aligning it with the center of each bar 
text(x = barplot(tweets_per_day$avg_tweets_per_days, plot = FALSE),
     y = tweets_per_day$avg_tweets_per_days,
     labels = round(tweets_per_day$avg_tweets_per_days, 2),
     pos = 3)

Plot for the average number of tweets per week for each HEI

barplot(tweets_per_week$avg_tweets_per_weeks,
        names.arg = tweets_per_week$id,
        main = "Average Tweets per Week",
        xlab = "HEI",
        ylab = "Average Number of Tweets",
        ylim = c(0, max(tweets_per_week$avg_tweets_per_weeks) + 5),
        las = 2,
        col = "#E74C3C")

text(x = barplot(tweets_per_week$avg_tweets_per_weeks, plot = FALSE),
     y = tweets_per_week$avg_tweets_per_weeks,
     labels = round(tweets_per_week$avg_tweets_per_weeks, 2),
     pos = 3)

Defining the intervals of time for the academic year

intervals <- list(
  interval1 = as.POSIXct(c("2022-08-31", "2022-12-15")),
  interval2 = as.POSIXct(c("2023-01-04", "2023-04-01")),
  interval3 = as.POSIXct(c("2023-04-14", "2023-06-15"))
)

Function to check if a date falls within a given interval of time and apply appropriate Boolean

check_interval <- function(date) {
  for (i in 1:length(intervals)) {
    interval_start <- intervals[[i]][1]
    interval_end <- intervals[[i]][2]
    if (date >= interval_start & date <= interval_end) {
      return(TRUE)
    }
  }
  return(FALSE)
}

data_tweets$academic_year <- sapply(data_tweets$created_at, check_interval)
print(data.frame(id = data_tweets$id, academic_year = data_tweets$academic_year))

Plot for the number of tweets per timeframe of either vacation or academic time

barplot(table(data_tweets$academic_year),
        main = "Number of Tweets per Timeframe",
        xlab = "Time",
        ylab = "Count",
        ylim = c(0, max(table(data_tweets$academic_year)) + 1000),
        names.arg = c("Vacation", "Academic"),
        col = c("#8E44AD", "#F1C40F"))

text(x = barplot(data_tweets$academic_year, plot = FALSE), 
     y = table(data_tweets$academic_year) + 0.5, 
     labels = table(data_tweets$academic_year), 
     pos = 3)

Function to count number of tweets and average per day

analyze_tweets <- function(academic_year_filter = TRUE) {
  # Filtering the data based on the academic_year_filter
  filtered_data <- data_tweets %>%
    filter(academic_year == academic_year_filter)
  
  # Count of days for each HEI
  unique_days <- filtered_data %>%
    group_by(id) %>%
    summarise(unique_days = n_distinct(as.Date(created_at)))
  
  # Count of tweets for each HEI
  number_tweets_boolean <- filtered_data %>%
    group_by(id) %>%
    summarise(count = n())
  
  # Naming the column respecting the time period
  year <- ifelse(academic_year_filter, "academic_time", "vacation_time")
  column_name <- paste0("avg_tweets_in_", year)
  
  # Combination of data and calculation of average posts per day
  combined_data <- left_join(unique_days, number_tweets_boolean, by = "id")
  combined_data <- combined_data %>%
    mutate(!!column_name := count / unique_days)
  
  print(combined_data)
  return(combined_data)
}

data_tweets_academic <- analyze_tweets()

data_tweets_vacations <- analyze_tweets(academic_year_filter = FALSE)

Plot for the average number of tweets during academic time for each HEI

barplot(data_tweets_academic$avg_tweets_in_academic_time,
        names.arg = data_tweets_academic$id,
        main = "Average Tweets during Academic Time",
        xlab = "HEI",
        ylab = "Average Number of Tweets",
        ylim = c(0, max(data_tweets_academic$avg_tweets_in_academic_time) + 5),
        las = 2,
        col = "#34495E")

text(x = barplot(data_tweets_academic$avg_tweets_in_academic_time, plot = FALSE),
     y = data_tweets_academic$avg_tweets_in_academic_time,
     labels = round(data_tweets_academic$avg_tweets_in_academic_time, 2),
     pos = 3)

Plot for the average number of tweets during vacation time for each HEI

barplot(data_tweets_vacations$avg_tweets_in_vacation_time,
        names.arg = data_tweets_vacations$id,
        main = "Average Tweets during Vacation Time",
        xlab = "HEI",
        ylab = "Average Number of Tweets",
        ylim = c(0, max(data_tweets_vacations$avg_tweets_in_vacation_time) + 5),
        las = 2,
        col = "#D35400")

text(x = barplot(data_tweets_vacations$avg_tweets_in_vacation_time, plot = FALSE),
     y = data_tweets_vacations$avg_tweets_in_vacation_time,
     labels = round(data_tweets_vacations$avg_tweets_in_vacation_time, 2),
     pos = 3)

Data preparation for day of the week

# Creating new table that contains a new column for the day of the week
data_tweets_days <- data_tweets %>%
  mutate(day_of_week = weekdays(created_at))

# Selecting only the id, created_at, and day_of_week columns for the new table
data_tweets_days <- data_tweets_days %>%
  select(id, created_at, day_of_week)

print(data_tweets_days)

# Grouping by id and day_of_week, then counting the number of tweets
number_tweets_days <- data_tweets_days %>%
  group_by(id, day_of_week) %>%
  summarise(count = n())

`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.

# Grouping by id, day_of_week and day created at, then counting the number of tweets
number_tweets_per_day <- data_tweets_days %>%
  mutate(created_date = as.Date(created_at)) %>%
  group_by(id, day_of_week, created_date) %>%
  summarise(count = n())

`summarise()` has grouped output by 'id', 'day_of_week'. You can override using the `.groups` argument.

# Finding for each HEI the average count of tweets per day
average_number_tweets_per_day <- number_tweets_per_day %>%
  group_by(id, day_of_week) %>%
  summarise(average_count = mean(count))

`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.

print(number_tweets_days)

Highest and lowest tweets

# Finding the HEI with the lowest count of tweets per day
lowest_count <- number_tweets_days %>%
  group_by(day_of_week) %>%
  slice_min(order_by = count) %>%
  select(day_of_week, id, count)

# Finding the HEI with the highest count of tweets per day
highest_count <- number_tweets_days %>%
  group_by(day_of_week) %>%
  slice_max(order_by = count) %>%
  select(day_of_week, id, count)

# Combine the results
high_low_HEI <- bind_rows(lowest_count, highest_count) %>%
  arrange(day_of_week)

print(high_low_HEI)

Plot for the lowest and highest count of tweets per day for each day of the week

ggplot(high_low_HEI, aes(x = day_of_week, y = count, fill = id)) +
  geom_bar(stat = "identity", position = "dodge") +
  geom_text(aes(label = count),
            position = position_dodge(width = 0.9),
            vjust = -0.5,
            size = 3) +
  labs(title = "Lowest and Highest Count of Tweets per Day for Each Day of the Week",
       x = "Day of the Week", y = "Count") +
  scale_fill_manual(values = rainbow(length(unique(high_low_HEI$id)))) +
  theme_minimal() +
  theme(legend.title = element_blank())

Average of tweets

# Finding the HEI with lowest and highest averaged count of tweets per day
high_low_average_HEIs <- average_number_tweets_per_day %>%
  group_by(day_of_week) %>%
  filter(average_count == max(average_count) | average_count == min(average_count)) %>%
  arrange(day_of_week, ifelse(average_count == min(average_count), average_count, -average_count))

print(high_low_average_HEIs)

Plot for the highest and lowest average count of tweets per day for each day of the week

ggplot(high_low_average_HEIs, aes(x = day_of_week, y = average_count, fill = id)) +
  geom_bar(stat = "identity", position = "dodge") +
  geom_text(aes(label = round(average_count, 2)),
            position = position_dodge(width = 0.7),
            vjust = -0.5,
            size = 3) +
  labs(title = "Highest and Lowest Average Count of Tweets per Day for Each Day of the Week",
       x = "Day of the Week", y = "Average Count") +
  scale_fill_manual(values = rainbow(length(unique(high_low_HEI$id)))) +
  theme_minimal() +
  theme(legend.title = element_blank())

Views Likes Retweets and Replies

# Table containing views, likes, retweets and replies for each media type for each HEI
types_of_tweets <- data_tweets %>%
              group_by(id, media_type) %>%
              summarise(count = n(),
                        views = sum(view_count, na.rm = TRUE),
                        likes = sum(favorite_count, na.rm = TRUE),
                        retweets = sum(retweet_count, na.rm = TRUE),
                        replies = sum(reply_count, na.rm = TRUE))

`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.

                        
print(types_of_tweets)                        

# Grouping by HEI and calculating the total values of views, likes and replies across all media types
total_tweets_stats <- types_of_tweets %>%
  group_by(id) %>%
  summarise(total_views = sum(views),
            total_likes = sum(likes),
            total_replies = sum(replies))

print(total_tweets_stats)

Function for piechart creation for views, likes and replies

pie_maker <- function(target_id = "duke.csv"){
  # Filtering data for the specific HEI
  hei_data <- types_of_tweets %>%
    filter(id == target_id)
  
  # Calculating total views for each media type for the specific HEI
  hei_media <- hei_data %>%
    group_by(media_type) %>%
    summarise(total_views = sum(views),
              total_likes = sum(likes),
              total_replies = sum(replies))
  
  # Calculating the percentage of views for each media type for the specific HEI
  hei_media$percentage_view <- hei_media$total_views / sum(hei_media$total_views) * 100
  hei_media$percentage_like <- hei_media$total_likes / sum(hei_media$total_likes) * 100
  hei_media$percentage_reply <- hei_media$total_replies / sum(hei_media$total_replies) * 100
  
  # Creating the pie chart for views
  hei_pie_chart_views <- ggplot(hei_media, aes(x = "", y = percentage_view, fill = media_type)) +
    geom_bar(stat = "identity", width = 1) +
    coord_polar("y", start = 0) +
    theme_void() +
    theme(legend.position = "right") +
    geom_text(aes(label = paste(media_type, "\n", total_views, "(", round(percentage_view, 1), "%)")), position = position_stack(vjust = 0.5), color = "#FFFFFF") +
    scale_fill_manual(values = c("no_media" = "#2196F3", "animated_gif" = "#E67E22", "photo" = "#8E44AD", "video" = "#138D75")) +
    labs(title = paste("Views for each media type -", target_id))
  
  # Creating the pie chart for likes
  hei_pie_chart_likes <- ggplot(hei_media, aes(x = "", y = percentage_like, fill = media_type)) +
    geom_bar(stat = "identity", width = 1) +
    coord_polar("y", start = 0) +
    theme_void() +
    theme(legend.position = "right") +
    geom_text(aes(label = paste(media_type, "\n", total_likes, "(", round(percentage_like, 1), "%)")), position = position_stack(vjust = 0.5), color = "#FFFFFF") +
    scale_fill_manual(values = c("no_media" = "#E91E63", "animated_gif" = "#4A148C", "photo" = "#90CAF9", "video" = "#00BFA5")) +
    labs(title = paste("Likes for each media type -", target_id))
  
  # Creating the pie chart for replies
  hei_pie_chart_replies <- ggplot(hei_media, aes(x = "", y = percentage_reply, fill = media_type)) +
    geom_bar(stat = "identity", width = 1) +
    coord_polar("y", start = 0) +
    theme_void() +
    theme(legend.position = "right") +
    geom_text(aes(label = paste(media_type, "\n", total_replies, "(", round(percentage_reply, 1), "%)")), position = position_stack(vjust = 0.5), color = "#FFFFFF") +
    scale_fill_manual(values = c("no_media" = "#666600", "animated_gif" = "#99CCCC", "photo" = "#9966CC", "video" = "#330000")) +
    labs(title = paste("Replies for each media type -", target_id))
  
  # Print the pie charts
  print(hei_pie_chart_views)
  print(hei_pie_chart_likes)
  print(hei_pie_chart_replies)
}

Plot of piecharts for each HEI

pie_maker()

pie_maker("epfl.csv")

pie_maker("goe.csv")

pie_maker("harvard.csv")

pie_maker("leicester.csv")

pie_maker("manchester.csv")

pie_maker("mit.csv")

pie_maker("sb.csv")

pie_maker("stanford.csv")

pie_maker("trinity.csv")

pie_maker("wv.csv")

pie_maker("yale.csv")

# Calculation of like_ratio and replies_ratio percentages
ratios_tweets_table <- total_tweets_stats %>%
  mutate(like_ratio = total_likes / total_views * 100,
         replies_ratio = total_replies / total_views * 100)

# Creation of new table with each HEI, like_ratio, and replies_ratio 
hei_tweets_ratios <- ratios_tweets_table %>%
  select(id, like_ratio, replies_ratio) %>%
  distinct()

print(hei_tweets_ratios)

Plot for like_ratio and replies_ratio for each HEI

ggplot(hei_tweets_ratios, aes(x = id)) +
  geom_bar(aes(y = like_ratio, fill = "Like Ratio"), stat = "identity", position = "dodge") +
  geom_bar(aes(y = replies_ratio, fill = "Replys Ratio"), stat = "identity", position = "dodge") +
  geom_text(aes(y = like_ratio, label = round(like_ratio, 2)), vjust = -0.5, position = position_dodge(width = 0.9), size = 3, color = "#000000") +
  geom_text(aes(y = replies_ratio, label = round(replies_ratio, 2)), vjust = -0.5, position = position_dodge(width = 0.9), size = 3, color = "#FFFFFF") +
  labs(title = "Like and Replys Ratios by HEI",
       x = "HEI",
       y = "Ratio (%)",
       fill = "Metric") +
  scale_fill_manual(values = c("Like Ratio" = "#2196F3", "Replys Ratio" = "#F44336")) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

# Table with averages of views, likes, retweets and replies
types_of_tweets_per_tweet <- types_of_tweets %>%
                        group_by(id, media_type) %>%
                        summarise(avg_views = mean(views / count),
                                  avg_likes = mean(likes / count),
                                  avg_retweets = mean(retweets / count),
                                  avg_replies = mean(replies / count))

`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.

print(types_of_tweets_per_tweet)

# Grouping by HEI and calculating the average values of views, likes and replies across all media types
total_average_stats <- types_of_tweets_per_tweet %>%
  group_by(id) %>%
  summarise(avg_views = sum(avg_views),
            avg_likes = sum(avg_likes),
            avg_replies = sum(avg_replies))

print(total_average_stats)

# Calculation of like_ratio and replies_ratio percentages
ratios_average_table <- total_average_stats %>%
  mutate(like_ratio = avg_likes / avg_views * 100,
         replies_ratio = avg_replies / avg_views * 100)

# Creation of new table with each HEI, like_ratio, and replies_ratio 
hei_average_ratios <- ratios_average_table %>%
  select(id, like_ratio, replies_ratio) %>%
  distinct()

print(hei_average_ratios)

Plot for like_ratio and replies_ratio for each HEI

ggplot(hei_average_ratios, aes(x = id)) +
  geom_bar(aes(y = like_ratio, fill = "Like Ratio"), stat = "identity", position = "dodge") +
  geom_bar(aes(y = replies_ratio, fill = "Replies Ratio"), stat = "identity", position = "dodge") +
  geom_text(aes(y = like_ratio, label = round(like_ratio, 2)), vjust = -0.5, position = position_dodge(width = 0.9), size = 3, color = "#000000") +
  geom_text(aes(y = replies_ratio, label = round(replies_ratio, 2)), vjust = -0.5, position = position_dodge(width = 0.9), size = 3, color = "#FFFFFF") +
  labs(title = "Like and Replies Ratios by HEI",
       x = "HEI",
       y = "Ratio (%)",
       fill = "Metric") +
  scale_fill_manual(values = c("Like Ratio" = "#330066", "Replies Ratio" = "#FF6666")) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

Hashtags

# Table with number of unique hashtags
unique_hashtags <- data_tweets %>%
                group_by(id) %>%
                summarise(count = n(),
                          unique_hashtags = length(unique(hashtags)))

print(unique_hashtags)

Plot for the count of unique hashtags for each HEI

barplot(unique_hashtags$unique_hashtags,
        names.arg = unique_hashtags$id,
        main = "Unique Hashtags for Each HEI",
        xlab = "HEI",
        ylab = "Count of Unique Hashtags",
        ylim = c(0, max(unique_hashtags$unique_hashtags) + 50),
        las = 2,
        col= "#16A085")

text(x = barplot(unique_hashtags$unique_hashtags, plot = FALSE),
     y = unique_hashtags$unique_hashtags,
     labels = round(unique_hashtags$unique_hashtags, 2),
     pos = 3)

Heatmaps

# Create column hour from created_at
data_tweets_days$created_hour <- as.numeric(format(data_tweets_days$created_at, "%H"))

Function to plot heatmap for various HEIs

heatmap_maker <- function(target_id = "duke.csv"){
  # Filtering data for the specific HEI
  target_data <- data_tweets_days %>%
    filter(id == target_id)
  
  # Grouping by day of the week and hour, and counting the number of tweets
  tweet_counts <- target_data %>%
    group_by(day_of_week, created_hour) %>%
    summarise(num_tweets = n())
  
  # Plotting heatmap
  ggplot(tweet_counts, aes(x = day_of_week, y = created_hour, fill = num_tweets)) +
    geom_tile() +
    scale_fill_gradient(low = "white", high = "blue") +
    labs(title = paste("Tweet Heatmap for", target_id),
         x = "Day of the week",
         y = "Hour of the day")
}

heatmap_maker()

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("epfl.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("goe.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("harvard.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("leicester.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("manchester.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("mit.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("sb.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("stanford.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("trinity.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("wv.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

heatmap_maker("yale.csv")

`summarise()` has grouped output by 'day_of_week'. You can override using the `.groups` argument.

Text

data_tweets_content <- data_tweets %>%
            select(id, text)

# Counting number of words
data_tweets_content <- data_tweets_content %>%
  mutate(num_words = lengths(strsplit(text, "\\s+")))

print(data_tweets_content)

# Grouping by HEI and calculate average, minimum, and maximum values of number of words
data_tweets_content_metrics <- data_tweets_content %>%
  group_by(id) %>%
  summarise(average_num_words = mean(num_words),
            min_num_words = min(num_words),
            max_num_words = max(num_words))
print(data_tweets_content_metrics)

Plot for the average, maximum and minimum values of words for each HEI

ggplot(data_tweets_content_metrics, aes(x = id, y = average_num_words)) +
  geom_point(aes(color = "Average")) +
  geom_errorbar(aes(ymin = min_num_words, ymax = max_num_words, color = "Range"), width = 0.2) +
  scale_color_manual(values = c("Average" = "#1976D2", "Range" = "#EF5350")) +
  labs(title = "Word Count Summary by HEI",
       x = "HEI",
       y = "Number of Words",
       color = "Metric") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

Now replies

data_replies <- data[data$type == "Reply", ]

data_replies

Interactions to replies

# Table containing views, likes, retweets and replies for each media type for each HEI
types_of_replies <- data_replies %>%
              group_by(id, media_type) %>%
              summarise(count = n(),
                        views = sum(view_count, na.rm = TRUE),
                        likes = sum(favorite_count, na.rm = TRUE),
                        retweets = sum(retweet_count, na.rm = TRUE),
                        replies = sum(reply_count, na.rm = TRUE))

`summarise()` has grouped output by 'id'. You can override using the `.groups` argument.

                        
print(types_of_replies)                        

# Grouping by HEI and calculating the total values of views, likes and replies across all media types
total_replies_stats <- types_of_replies %>%
  group_by(id) %>%
  summarise(total_views = sum(views),
            total_likes = sum(likes),
            total_replies = sum(replies))

print(total_replies_stats)

# Calculation of like_ratio and replies_ratio percentages
ratios_replies_table <- total_replies_stats %>%
  mutate(like_ratio = total_likes / total_views * 100,
         replies_ratio = total_replies / total_views * 100)

# Creation of new table with each HEI, like_ratio, and replies_ratio 
hei_replies_ratios <- ratios_replies_table %>%
  select(id, like_ratio, replies_ratio) %>%
  distinct()

print(hei_replies_ratios)

Clusters

# Creating table for cluster algorithms

# Joining attribute count (number of tweets) and unique_hashtags (number of unique hashtags) per HEI
cluster_table <- merge(select(unique_hashtags, id, unique_hashtags), select(number_tweets, id, count), by = "id", all=TRUE)

# Joining attribute avg_tweets_per_days (average of tweets per day) per HEI
cluster_table <- merge(cluster_table, select(tweets_per_day, id, avg_tweets_per_days), by = "id", all=TRUE)

# Joining attribute avg_tweets_per_weeks (average of tweets per week) per HEI
cluster_table <- merge(cluster_table, select(tweets_per_week, id, avg_tweets_per_weeks), by = "id", all=TRUE)

# Joining attribute avg_tweets_in_academic_time (average of tweets during academic time) per HEI
cluster_table <- merge(cluster_table, select(data_tweets_academic, id, avg_tweets_in_academic_time), by = "id", all=TRUE)

# Joining attribute avg_tweets_in_vacation_time (average of tweets during vacation time) per HEI
cluster_table <- merge(cluster_table, select(data_tweets_vacations, id, avg_tweets_in_vacation_time), by = "id", all=TRUE)

# Joining attribute total_views (total number of views), total_likes (total number of likes) and total_replies (total number of replies) per HEI
cluster_table <- merge(cluster_table, select(total_tweets_stats, id, total_views, total_likes, total_replies), by = "id", all=TRUE)

# Renaming attribute like_ratio to total_like_ratio and replies_ratio to total_replies_ratio
cluster_table <- merge(cluster_table, select(hei_tweets_ratios, id, like_ratio, replies_ratio), by = "id", all=TRUE)
cluster_table <- cluster_table %>%
  rename(total_like_ratio = like_ratio, 
         total_replies_ratio = replies_ratio)

# Joining attribute avg_views (average number of views), avg_likes (average number of likes) and avg_replies (average number of replies) per HEI
cluster_table <- merge(cluster_table, select(total_average_stats, id, avg_views, avg_likes, avg_replies), by = "id", all=TRUE)

# Renaming attribute like_ratio to avg_like_ratio and replies_ratio to avg_replies_ratio
cluster_table <- merge(cluster_table, select(hei_average_ratios, id, like_ratio, replies_ratio), by = "id", all=TRUE)
cluster_table <- cluster_table %>%
  rename(avg_like_ratio = like_ratio, 
         avg_replies_ratio = replies_ratio)

print(cluster_table)

Function for cluster method

cluster_maker <- function(seed = 123, num_clusters = 3, table){
  set.seed(123)
  
  # Excluding id column for clustering
  cluster_data <- select(table, -id)
  
  # Scaling the data for kmeans method
  scaled_data <- scale(cluster_data)
  
  kmeans_result <- kmeans(scaled_data, centers = num_clusters)
  
  print(kmeans_result$centers)
  print(kmeans_result$cluster)
  
  return(kmeans_result)
}

Function to add ids to better visualize results

cluster_id_maker <- function(kmeans_result, table){
  # Merging the cluster assignments with the original data
  cluster_assignments <- data.frame(id = table$id, cluster = kmeans_result$cluster)

  print(cluster_assignments)
  plot(kmeans_result$cluster)
}

Three clusters with seed 123

cluster_123_3 <- cluster_maker(table = cluster_table)

  unique_hashtags      count avg_tweets_per_days avg_tweets_per_weeks avg_tweets_in_academic_time avg_tweets_in_vacation_time total_views total_likes
1      -0.6314353 -0.9289798          -0.9124788           -0.9124788                  -0.8237303                  -0.8263586  -0.4559850  0.07640829
2      -0.4672956  1.2357198           1.2354314            1.2354314                   1.1911790                   1.2741657   2.0971401  1.93751241
3       0.1740030 -0.1713844          -0.1731538           -0.1731538                  -0.1731809                  -0.1913303  -0.4153661 -0.43904812
  total_replies total_like_ratio total_replies_ratio  avg_views  avg_likes avg_replies avg_like_ratio avg_replies_ratio
1     0.5534067        2.0996788           3.0684868 -0.3467066  1.9304181   2.3142417      2.6030420         3.1020546
2     1.6810073       -0.7177677          -0.4777518  1.7460208  1.4388566   1.1599670     -0.6396304        -0.4169707
3    -0.4350468       -0.0737937          -0.2347759 -0.3494817 -0.5342368  -0.5149084     -0.1470868        -0.2520126
 [1] 3 3 3 2 3 3 2 3 1 3 3 3

cluster_id_maker(cluster_123_3, table = cluster_table)

Seven clusters with seed 123

cluster_123_6 <- cluster_maker(num_clusters = 7, table = cluster_table)

  unique_hashtags      count avg_tweets_per_days avg_tweets_per_weeks avg_tweets_in_academic_time avg_tweets_in_vacation_time total_views total_likes
1     -0.54099099 -0.7980288          -0.7998764           -0.7998764                  -0.8566746                  -0.7686479  -0.4659280 -0.34870651
2     -0.33665384  2.0980363           2.1001655            2.1001655                   2.2152351                   2.2141794   1.6390693  2.76526795
3      2.06849496  0.3163906           0.3164848            0.3164848                   0.3487604                   0.3235393  -0.4385686 -0.45795609
4     -0.36345215 -0.5258824          -0.5295450           -0.5295450                  -0.5074127                  -0.5879623  -0.3946334 -0.54854585
5     -0.59793741  0.3734033           0.3706973            0.3706973                   0.1671229                   0.3341520   2.5552108  1.10975687
6     -0.63143530 -0.9289798          -0.9124788           -0.9124788                  -0.8237303                  -0.8263586  -0.4559850  0.07640829
7     -0.03517279  1.5243464           1.5265794            1.5265794                   1.4868514                   1.5200937  -0.3507682 -0.14392452
  total_replies total_like_ratio total_replies_ratio  avg_views  avg_likes avg_replies avg_like_ratio avg_replies_ratio
1   -0.44336928        1.2911568          0.07105937 -0.3861273 -0.3579270  -0.4728730      0.8212504       -0.10327394
2    2.91587745       -0.6417247         -0.40143781  0.3936984  1.4654190   1.5919118     -0.5339425       -0.30190583
3   -0.46641612       -0.2251792         -0.13401717 -0.3929316 -0.6804893  -0.5991306     -0.2493630       -0.12958214
4   -0.51767840       -0.6173697         -0.47022920 -0.3036546 -0.5504837  -0.5144803     -0.5587944       -0.43018681
5    0.44613716       -0.7938108         -0.55406573  3.0983433  1.4122942   0.7280222     -0.7453183       -0.53203567
6    0.55340665        2.0996788          3.06848680 -0.3467066  1.9304181   2.3142417      2.6030420        3.10205460
7   -0.02513687       -0.3266195         -0.10615086 -0.3725990 -0.5293640  -0.4322473     -0.2323782       -0.08165369
 [1] 4 4 1 2 3 4 5 4 6 3 1 7

cluster_id_maker(cluster_123_6, table = cluster_table)

Five clusters with seed 123

cluster_123_6 <- cluster_maker(num_clusters = 5, table = cluster_table)

  unique_hashtags      count avg_tweets_per_days avg_tweets_per_weeks avg_tweets_in_academic_time avg_tweets_in_vacation_time total_views total_likes
1      -0.6314353 -0.9289798          -0.9124788           -0.9124788                  -0.8237303                  -0.8263586  -0.4559850  0.07640829
2      -0.3366538  2.0980363           2.1001655            2.1001655                   2.2152351                   2.2141794   1.6390693  2.76526795
3       1.3672724  0.7190425           0.7198496            0.7198496                   0.7281241                   0.7223908  -0.4093018 -0.35327890
4      -0.4226318 -0.6165979          -0.6196555           -0.6196555                  -0.6238333                  -0.6481908  -0.4183983 -0.48193274
5      -0.5979374  0.3734033           0.3706973            0.3706973                   0.1671229                   0.3341520   2.5552108  1.10975687
  total_replies total_like_ratio total_replies_ratio  avg_views  avg_likes avg_replies avg_like_ratio avg_replies_ratio
1     0.5534067       2.09967875           3.0684868 -0.3467066  1.9304181   2.3142417      2.6030420         3.1020546
2     2.9158774      -0.64172468          -0.4014378  0.3936984  1.4654190   1.5919118     -0.5339425        -0.3019058
3    -0.3193230      -0.25899262          -0.1247284 -0.3861541 -0.6301142  -0.5435029     -0.2437014        -0.1136060
4    -0.4929087       0.01880576          -0.2897997 -0.3311455 -0.4862981  -0.5006112     -0.0987795        -0.3212159
5     0.4461372      -0.79381078          -0.5540657  3.0983433  1.4122942   0.7280222     -0.7453183        -0.5320357
 [1] 4 4 4 2 3 4 5 4 1 3 4 3

cluster_id_maker(cluster_123_6, table = cluster_table)

Four clusters with seed 4855

cluster_123_3 <- cluster_maker(seed = 4855, num_clusters = 4, table = cluster_table)

  unique_hashtags      count avg_tweets_per_days avg_tweets_per_weeks avg_tweets_in_academic_time avg_tweets_in_vacation_time total_views total_likes
1      -0.6314353 -0.9289798          -0.9124788           -0.9124788                  -0.8237303                  -0.8263586  -0.4559850  0.07640829
2      -0.4672956  1.2357198           1.2354314            1.2354314                   1.1911790                   1.2741657   2.0971401  1.93751241
3       1.3672724  0.7190425           0.7198496            0.7198496                   0.7281241                   0.7223908  -0.4093018 -0.35327890
4      -0.4226318 -0.6165979          -0.6196555           -0.6196555                  -0.6238333                  -0.6481908  -0.4183983 -0.48193274
  total_replies total_like_ratio total_replies_ratio  avg_views  avg_likes avg_replies avg_like_ratio avg_replies_ratio
1     0.5534067       2.09967875           3.0684868 -0.3467066  1.9304181   2.3142417      2.6030420         3.1020546
2     1.6810073      -0.71776773          -0.4777518  1.7460208  1.4388566   1.1599670     -0.6396304        -0.4169707
3    -0.3193230      -0.25899262          -0.1247284 -0.3861541 -0.6301142  -0.5435029     -0.2437014        -0.1136060
4    -0.4929087       0.01880576          -0.2897997 -0.3311455 -0.4862981  -0.5006112     -0.0987795        -0.3212159
 [1] 4 4 4 2 3 4 2 4 1 3 4 3

cluster_id_maker(cluster_123_3, table = cluster_table)

Six clusters with seed 4855

cluster_123_6 <- cluster_maker(seed = 4855, num_clusters = 6, table = cluster_table)

  unique_hashtags      count avg_tweets_per_days avg_tweets_per_weeks avg_tweets_in_academic_time avg_tweets_in_vacation_time total_views total_likes
1      -0.5409910 -0.7980288          -0.7998764           -0.7998764                  -0.8566746                  -0.7686479  -0.4659280 -0.34870651
2      -0.3366538  2.0980363           2.1001655            2.1001655                   2.2152351                   2.2141794   1.6390693  2.76526795
3       1.3672724  0.7190425           0.7198496            0.7198496                   0.7281241                   0.7223908  -0.4093018 -0.35327890
4      -0.3634522 -0.5258824          -0.5295450           -0.5295450                  -0.5074127                  -0.5879623  -0.3946334 -0.54854585
5      -0.5979374  0.3734033           0.3706973            0.3706973                   0.1671229                   0.3341520   2.5552108  1.10975687
6      -0.6314353 -0.9289798          -0.9124788           -0.9124788                  -0.8237303                  -0.8263586  -0.4559850  0.07640829
  total_replies total_like_ratio total_replies_ratio  avg_views  avg_likes avg_replies avg_like_ratio avg_replies_ratio
1    -0.4433693        1.2911568          0.07105937 -0.3861273 -0.3579270  -0.4728730      0.8212504        -0.1032739
2     2.9158774       -0.6417247         -0.40143781  0.3936984  1.4654190   1.5919118     -0.5339425        -0.3019058
3    -0.3193230       -0.2589926         -0.12472840 -0.3861541 -0.6301142  -0.5435029     -0.2437014        -0.1136060
4    -0.5176784       -0.6173697         -0.47022920 -0.3036546 -0.5504837  -0.5144803     -0.5587944        -0.4301868
5     0.4461372       -0.7938108         -0.55406573  3.0983433  1.4122942   0.7280222     -0.7453183        -0.5320357
6     0.5534067        2.0996788          3.06848680 -0.3467066  1.9304181   2.3142417      2.6030420         3.1020546
 [1] 4 4 1 2 3 4 5 4 6 3 1 3

cluster_id_maker(cluster_123_6, table = cluster_table)